Improved Quantification for PET/MR
Positron emission tomography (PET) allows for the quantitative measurement of the distribution of a radioactive tracer within the human body in vivo. However, one major challenge for accurate quantification in PET/MR is attenuation correction (AC), which is necessary to compensate for the effect of photon attenuation within the patient. Unlike for PET/CT, where attenuation values can be directly derived from the CT images, AC for PET/MR is challenging. This is because the MR images reflect the spatial distribution of water and fat protons and thus do not provide direct information on the attenuating properties of the underlying tissue. Standard MR-based AC (MR-AC) acquires a series of dedicated MR images, which are used to segment several tissue classes with pre-defined attenuation values. However, MR-AC does not account for the presence of bone and thus leads to an underestimation of the activity distribution, especially in close vicinity to bone.
To improve quantification for PET/MR, we are developing algorithms which simultaneously reconstruct activity and attenuation distributions from the PET emission data using available MR images as anatomical prior information. The MR information is used to derive voxel-dependent expectations on the attenuation coefficients. An iterative reconstruction scheme incorporating the prior information on the attenuation coefficients is used to update attenuation and activity distribution in an alternating manner. The proposed methods can recover bone attenuation information and therefore improve PET quantification compared to MR-AC.
Figure 1: Attenuation map and activity difference map of standard MR-based AC and one of our proposed methods compared to the ground truth.
Another major obstacle to accurate quantification in PET/MR is involuntary patient motion during measurements, such as respiration, cardiac motion or muscle relaxation. It leads to image blurring and, in case of PET, to an underestimation of the reconstructed activity. A widely used motion handling strategy is gating, which typically represents a trade-off between good temporal resolution and an appropriate signal-to-noise ratio and contrast-to-noise ratio of the reconstructed data. Since the advent of fully-integrated systems combining PET and magnetic resonance (MR), several new approaches for motion handling have been proposed, which derive motion information from MR measurements. Typically, a non-rigid registration algorithm is applied for estimating motion vector fields, which describe patient motion from phase to phase and allow for motion compensation of PET images.
We are developing new methods for respiratory motion compensation of PET images using information from a strongly undersampled radial MR sequence that runs in parallel with the PET acquisition, can be interlaced with clinical MR sequences, and requires less than one minute of the total MR acquisition time per bed position. This new method yields either reduced motion blurring of PET images and improved quantification accuracy of radiotracer distribution or lower noise levels of PET images, thus, increasing the diagnostic value of PET/MR imaging.
Figure 2: Comparison of reconstructions of simulated PET data: 3D, 4D gated, and 4D motion-compensated reconstruction employing one of our newly-developed methods.
© dkfz.de